HLTH 511 Week 3 Discussion Board Forum 3 Example

Discussion Board Forum 3: Study Design and Causal Inference

HLTH 511 Research Methods in Public Health, Discussion Board Forum 3

Thread by [Student Name]

The classmates are composites; the textbook, Scripture and the peer-reviewed sources cited are real.

For years, many physicians recommended hormone therapy to postmenopausal women partly because it seemed to protect the heart. That advice rested largely on observational studies, and it changed dramatically when a randomized trial reported its results. The reversal affected millions of women and remains a standard case in epidemiology courses, because it shows how a careful, respected study can still point in the wrong direction.

Jacobsen (2017) distinguishes observational studies, in which researchers measure exposures that participants choose for themselves, from experimental studies, in which researchers assign exposures, ideally at random. Randomization balances both measured and unmeasured characteristics between groups, so differences in outcomes can be attributed to the exposure. Observational studies must instead adjust for confounders, and they can only adjust for factors they measure. Cohort studies have real strengths, including large samples, long follow-up and the ability to study many outcomes, but their conclusions about cause depend on how completely confounding has been controlled.

Grodstein et al. (1996) followed 59,337 women in the Nurses' Health Study for up to 16 years and found that women taking estrogen with progestin had a multivariate adjusted relative risk of 0.39 for major coronary heart disease compared with women not using hormones, a 61% reduction. The Women's Health Initiative then randomized 16,608 postmenopausal women to estrogen plus progestin or placebo (Writing Group for the Women's Health Initiative Investigators, 2002). After about five years, the trial was stopped early: women on hormones had 1.29 times the hazard of coronary heart disease, 1.41 times the hazard of stroke, for pulmonary embolism 2.13 and for invasive breast cancer 1.26, while hip fractures and colorectal cancer were less common.

The most widely accepted explanation is confounding by the healthy user effect. Women who chose hormone therapy in the cohort tended to be wealthier, more educated, more likely to see a physician regularly and more health-conscious in general, and those traits independently lower heart disease risk. Statistical adjustment can account for the confounders researchers think to measure, but randomization also balances the ones nobody thought of. Differences in women's ages and the timing of hormone use may also have played a part, since trial participants tended to be older and more years past menopause than many cohort participants. For public health practice, the lesson is not to dismiss cohort studies but to treat a recommendation affecting millions of healthy people as a hypothesis to test before it becomes routine care.

Acts 17:11 commends the Bereans, who received Paul's message eagerly but examined the Scriptures daily to see if what they heard was true (English Standard Version Bible, 2001/2016). Public health professionals should welcome promising findings with the same openness and the same diligence, especially when millions of people may act on them. Testing a claim is not a sign of distrust; it is a way of caring for those who will live with the answer.

Reply to Hannah

Hannah, you asked whether better statistical adjustment could have prevented the hormone therapy error, and I think it is a fair question to explore, since adjustment is the main tool we will use in our own projects. Jacobsen (2017) explains that multiple regression allows researchers to estimate the association between an exposure and an outcome while holding other measured variables constant. The Nurses' Health Study did adjust for many confounders, including age, smoking, blood pressure and cholesterol. The difficulty is that regression handles only the variables a study records, and records accurately. Traits such as general health-consciousness or adherence to medical advice are hard to capture in a questionnaire, yet they influence both the decision to take hormones and the risk of heart disease. Grodstein et al. (1996) reported a relative risk of 0.39 even after adjustment, which shows how much bias can remain. Some researchers now use methods such as propensity scores or comparisons of new users to reduce this bias, but none can fully replace randomization. Your question is a good reminder that a very strong protective association in an observational study should invite more scrutiny, not less, because unusually large benefits are sometimes a sign that confounding has not been fully removed. In our own research projects, we will be adjusting for covariates in SPSS, and it will be worth stating plainly in our limitations which important confounders our data could not measure. Naming those gaps honestly helps readers judge how much weight our findings can bear.

Reply to Marcus

Marcus, you rightly and helpfully pointed out that many public health questions cannot be studied in randomized trials, such as the health effects of smoking, air pollution or poverty, since we cannot ethically assign people to harmful exposures. I agree, and Jacobsen (2017) notes that observational designs are often the only ethical or practical option, which is why so much of public health rests on them. The hormone therapy case does not mean observational studies are untrustworthy; it means we should weigh them carefully. When trials are impossible, researchers can strengthen observational evidence by looking for consistency across many studies and populations, a dose-response relationship, a plausible biological mechanism and natural experiments, such as policy changes that affect some communities but not others. The evidence that smoking causes lung cancer, for instance, came from observational studies that met all of these tests. The key lesson from that trial (Writing Group for the Women's Health Initiative Investigators, 2002) is that when a trial is possible and the stakes are high, it should be done before an intervention becomes standard practice. Applying that standard consistently would protect the public from both premature enthusiasm and needless skepticism about observational research. Many questions in our field, such as the effects of housing or neighborhood violence on health, will always depend on observational data, so learning to read those studies critically, rather than dismissing them, is one of the most useful skills this course can give us. Your examples make that point clearly and well.

References

English Standard Version Bible. (2016). Crossway. (Original work published 2001)

Grodstein, F., Stampfer, M. J., Manson, J. E., Colditz, G. A., Willett, W. C., Rosner, B., Speizer, F. E., & Hennekens, C. H. (1996). Postmenopausal estrogen and progestin use and the risk of cardiovascular disease. New England Journal of Medicine, 335(7), 453-461. https://doi.org/10.1056/NEJM199608153350701

Jacobsen, K. H. (2017). Introduction to health research methods: A practical guide (2nd ed.). Jones & Bartlett Learning.

Writing Group for the Women's Health Initiative Investigators. (2002). Risks and benefits of estrogen plus progestin in healthy postmenopausal women: Principal results from the Women's Health Initiative randomized controlled trial. JAMA, 288(3), 321-333. https://doi.org/10.1001/jama.288.3.321